The present invention relates to a method, apparatus, and computer program for processing digital images to estimate noise.
The color cross-correlation properties of photographic recording media have been used to produce enhanced renditions of digital images to formulate a color matrix transform used to enhance the appearance of scanned images. In U.S. Pat. No. 5,509,086 Edgar et al. disclose a method for correcting for the effects of crosscolor crosstalk between colors of a color image stored in an image storage medium and an imaging device used to scan the color image into a plurality of scanned images each corresponding to a color of the color image. The method Edgar et al. describe comprises the steps of a) identifying noise patterns in the plurality of scanned images; b) calculating crosscolor correlations between the identified noise patterns in the plurality of scanned images; and c) correcting for color crosstalk in each of the scanned images according to the crosscolor correlations. This method makes explicit use of measured color cross-correlation in digital images to formulate a color matrix transform to enhance the appearance of the scanned images.
The color cross-correlation properties of photographic recording media have been used to produce enhanced renditions of digital images based on applying a noise reduction algorithm to digital images. In U.S. Pat. No. 5,956,432, Ohta discloses an image processing apparatus designed to remove noise from a color digital image. The apparatus includes detection means for detecting the correlation between color-component signals of an input image signal, and a smoothing means for performing smoothing processing on the image signal inversely proportional to the correlation detected by the detection means.
Image structure content relating to natural scene objects exhibits a different cross-color correlation property than does the stochastic noise content resulting from the imaging medium. This cross-color correlation property has been used to advantage for the formation of a color transform as described by Edgar et al. and for the purposes of noise removal by as described by Ohta.
It is therefore an object of the present invention to provide an improved noise estimation algorithm that takes advantage of the cross-color correlation property of images, thus improving other image enhancement algorithm and that relies on prior knowledge of the noise characteristics of digital images.
This object is achieved by a method of calculating a noise characteristic value for a source digital image, comprising the steps of:
a) receiving a source digital image including pixels corresponding to two or more different colors;
b) using pixel values corresponding to a first color to produce at least one color correlation term for pixels corresponding to a second color which respectively indicate the extent of the noise content in the pixel values of the second color; and
c) using the color correlation term to calculate a noise characteristic value for the pixels corresponding to the second color.